Radio frequency (RF) energy is used in various industries for the treatment of materials through induction heating, dielectric heating, and plasma excitation. Plasma excitation can take the form of inductive, capacitive, or true electromagnetic (EM) wave, microwave, couplings. Generators which provide this RF energy utilize many circuit topologies ranging from single class A transistor amplifiers providing a few tens of watts to self-oscillating tube (valve) generators providing many thousands of watts.
The semiconductor manufacturing industry utilizes RF plasmas for depositing and etching micron and sub-micron sized films. A typical power supply for this application may consist of a line frequency transformer/rectifier/capacitor DC power supply and high frequency power amplifier. Typical power and frequency values may be up to 20 KW within the range of 400 KHz to 100.0 MHz, but may not be so limited. A power supply or generator would have power controllable to 1 or 2% precision over a 100:1 output load range. Usually the generator is specifically configured to output to a defined load, usually 50 ohms, but should be able to drive any load, even if mismatched, without failure. If a mismatch condition (e.g., increase in load impedance) and/or other condition arises such that current draw from the power amplifier and supplied to the load exceeds a certain threshold, damage can occur to the power amplifier over time. Typical protection schemes reduce the power. For example, the drive level to a linear amplifier is reduced to correspondingly reduce current or power dissipation. In a 50 ohm system, variation from the typical 50 ohms can be measured as reflected power. The drive level is reduced to limit reflected power.
FIG. 1 shows a typical transformer-coupled push-pull RF power amplifier having switches or transistors S1, S2 driven by sine waves which are out of phase. A five element harmonic rejection filter includes inductors L1, L2 and capacitors C1, C2, and C4. The harmonic rejection filter typically ensures a high purity or uniform sine wave output. No biasing schemes are shown which may be class AB or class B. Either bipolar junction transistors (BJTs) or metal oxide semiconductor field effect transistors (MOSFETs) are typically used. The transformer T1 has a ratio chosen to match the required power for a given DC supply voltage, usually 28V or 50V. Detailed circuitry follows standard industry practice for broadband HF/VHF power amplifier design as would be used for communications.
The amplifier of FIG. 1 offers one advantage, but several disadvantages. The advantage is that in a broadband design, the output frequency is easily changed simply by varying the drive or input frequency. For a given output frequency, only the output filter needs to be changed. If the basic linearity/purity of the amplifier is good enough, dispensed with altogether. The circuit of FIG. 1 has the disadvantages of poor efficiency and high transistor power dissipation. Efficiency theoretically cannot exceed 70% but typically is no better than 50%. To address the high power dissipation, many applications use expensive, special RF transistors which often employ beryllium oxide (BEo) low thermal resistance technology. This often requires large air or water cooled heatsinks. There is a large amount of data published on RF linear amplifier design. Any power supply manufacturer desiring to design a generator can use the transistor manufacturer's application circuit with a high degree of confidence.
As can be seen in FIG. 2, the circuit of FIG. 2 utilizes a different mode of operation offering high efficiency and low power dissipation. The drive signals in the circuit of FIG. 2 are fixed at square waves so that the transistors are now in a switching rather than a linear mode of operation. That is, the switches or transistors S1, S2 of FIG. 1 operate in a region between fully off and fully on. The switches or transistors S1, S2 of FIG. 2 operate by switching from fully on to fully off. The output of transformer T1 is now a square wave. A four element filter including inductors L1, L2 and capacitors C1, C2 filters out the required fundamental frequencies to yield a sinusoidal output. Capacitor C4 is removed so that the filter provides an inductive input, in order to reject harmonic current. Although the transistor and transformer voltages are square, the currents are sinusoidal. Efficiency can now be 100%, and typically falls within the range of 80-95%. Such a circuit is usually referred to as a resonant converter or inverter rather than an amplifier.
The circuit of FIG. 2 suffers some disadvantages. The filter is sufficiently selected for a particular output frequency so that only a fixed or narrow frequency range or band of operation is possible. Also, the output power cannot be directly controlled. Unlike, FIG. 1, the circuit of FIG. 2 cannot connect directly to a line or outlet voltage. Rather, the DC input to FIG. 2 requires regulation using an additional power converter, typically implemented using a switched mode converter. Further, mismatch loads can cause high circulating currents between the filter and transistors. The circulating currents are not necessarily limited by limiting the DC input current.
With particular respect to class E amplifiers, a class E amplifier is a switch-mode amplifier topology offering high efficiency. Because of its topology, the switch element, typically a transistor, of the class E amplifier spends little or no time in the active region where the greatest power dissipation occurs. In this configuration, the switch element of the class E amplifier operates more like a switch rather than a transistor. That is, the switching element spends the majority of its time in either the cutoff or the saturation regions.
Designers further improve the efficiency of the class E amplifier by using a switch-mode technique known as the zero-voltage switching (ZVS). ZVS prevents the switch element of the class E amplifier from passing through the active region during transitions. By applying an inductive load at the output of the switch element, the parasitic and swamping capacitances at the output of the switch element are discharged to zero volts before the switch element attempts to transition from the cutoff region to the saturation region. An inductor and a capacitor cooperate to form a series-resonant circuit and provide an inductive load at an output of the switch element. The frequency of the resonant circuit is less than the operating frequency of the amplifier. When this occurs, the inductor of the resonant circuit dominates the resonant circuit and generates an inductive load on the transistor.
In order to perform ZVS, the switch element must be designed to permit negative drain-source current to pass through it, even if the device channel is in the cutoff region. Such a requirement suggests that a MOSFET is a preferred selection for the switch element of the class E amplifier topology because MOSFETs have an intrinsic body diode at the substrate connection to the source. Other transistors may be selected, such as a bipolar junction transistor (BJT) or an integrated gate bipolar transistor (IGBT), but such a configuration requires that a fast diode be placed across the emitter-collector junction.
The primary benefit of the class E amplifier is that more RF power can be realized from the same transistor used in a class E topology versus other topologies, primarily due to reduced device dissipation. On the other hand, the class E amplifier generates substantial second harmonic energy that must be removed from the RF output. Such topologies typically require at least one additional stage of filtering before the RF power is delivered to the load.
As discussed previously, the series-resonant circuit consisting of an inductor and a capacitor possess a resonant frequency below the amplifier operating frequency. Although the load could be any combination of capacitors, inductors, and resistors, if the load is only a capacitor with a value such that the series combination of the resonant circuit and the load has a resonant frequency equal to the amplifier operating frequency, the current through the switch element could approach an infinite value. This could result in damage to the transistor. Typical class E amplifier applications, however, avoid transistor damage by utilizing an external control loop that clamps the amplifier output reflected power. Once the control loop senses that the reflected power has exceeded a preset limit, the control loop reduces the voltage at the DC rail until the reflected power matches a predetermined limit. The control loop must react quickly in order to avoid impact to the transistor. Impact to the transistor can also be avoided by reducing the RF amplifier input power to zero. However, in a plasma processing application, such an action may cause the undesirable result that the plasma is extinguished.